Many severely brain-injured patients have no reliable motor output channel, making accurate assessments of cognitive ability very difficult or impossible. Given this confound, it is essential that alternative approaches to the bedside exam be developed to allow for more accurate measurements of a patient's cognitive state. This proposal aims to use EEG measures of the neural response to auditory language for this purpose. Specifically, advanced spectral methods will be used to characterize temporal and spatial aspects of EEG responses to natural language in severely brain-injured subjects and controls. We will conduct a nested and hierarchical study of the brain's response to language. In particular, we will address three questions: is there a differential response to forward versus backward spoken language, does the brain respond differently to different categories of individual words, and does the brain respond to particular markers of semantic content in continuous narration. Analysis of these three levels of language content will allow us to build a graded representation of an individual subject's neural response to language. The central hypothesis is that markers of the EEG-measured response to language will elucidate the neural underpinnings of cognitive function after brain injury. Further, for those patients who do demonstrate consistent bedside behavior, I hypothesize that these measures will be linked to behavioral criteria operationally defined by the Coma Recovery Scale-Revised (CRS-R). First, in order to identify changes in the EEG that co-occur with presentation of language content, subjects will listen to personally relevant stories as well as to time-reversed versions of the same. Analysis of this data will reveal whether there is a specific pattern of brain activity that selectively reflects language processing, and whether such markers relate to outward behavior of patients. Second, to determine if subjects demonstrate a differential neural response to words from different categories, normal subjects and patients will listen to repetitions of ten nouns from two categories previously shown to be easily discriminated in normal brains: objects than can be manipulated by hand and objects than can provide shelter. I will then train a classifier to distinguish between the different categories of words. I will first demonstrate that he brain activity of normal subjects and at least some brain-injured subjects shows a differential response that can be classified. We will then test the hypothesis that the success of a classifier trained on patient data will grade with the patient's CRS-R score. Third, to identify neural activity that occurs in response to narrative shifts in continuous narration, naturalistic narrativs will be played to normal subjects and severely brain-injured subjects during continuous EEG. The data around narrative shifts will be analyzed using average spectra and spectrograms around the events of interest to determine whether there is evidence that subjects are able to actively follow a story. PUBLIC HEALTH RELEVANCE: My central hypothesis is that markers of the EEG-measured response to language will elucidate the neural underpinnings of cognitive function after severe traumatic brain injury. Further, for patients who demonstrate consistent bedside behavior, I hypothesize that these measures will be linked to operationally defined behavioral criteria. As such, the proposed experiments could provide a new set of tools for assaying the cognitive function of a brain-injured subject whose diagnosis is otherwise uncertain.